CMOL Circuits (“CrossNets”)
The CMOL circuit fabric is uniquely suitable for the implementation of neuromorphic networks (“CrossNets”) in which cell somas are realized the CMOS subsystem, crossbar nanowires play the roles of axons and dendrites, and crosspoint latching switches serve as elementary (binary-weight) synapses. The important advantage of this topology is the possibility to implement arbitrary cell connectivity (e.g., ~104 typical for the mammal cortex) in quasi-2D electronic circuits. We have shown that the binary character of the elementary synapses and a relatively high defect density (possible at the initial stage of CMOL technology development) do not prevent CrossNets from performing essentially all the tasks demonstrated earlier with software-implemented neuromorphic networks, including auto-association , pattern classification [9-11, 14], and dynamic control in conditions of instant and delayed reward [12, 13]. The significance of these results is in the very high potential areal density of CMOL CrossNets (beyond that of the mammal cerebral cortex, at similar connectivity), and the very high operation speed of these networks – e.g., intercell latency below 1 microsecond at readily manageable power dissipation below 1 W/cm2 [5, 8]. We believe that CMOL CrossNets is the first hardware which may eventually challenge the human cortex. At a shorter time scale, such circuits may become an important tool for cortical circuit modeling.
1. S. Fölling, Ö. Türel, and K. K. Likharev, "Single-Electron Latching Switches as Nanoscale Synapses", in: Proc. IJCNN’01Neural Networks, pp. 216-221 (2001).
2. Ö. Türel and K. K. Likharev, "CrossNets: Possible Neuromorphic Networks based on Nanoscale Components", Int. J. of Circuit Theory and Applications 31, pp. 37-52 (2003).
3. Ö. Türel and K. K. Likharev, "CrossNets: Neuromorphic Networks for Nanoelectronic Implementation", Lecture Notes on Computer Science 2714, pp. 753-760 (2003).
4. Ö. Türel,
5. K. Likharev, A. Mayr,
6. Ö. Türel, J. H. Lee, X. Ma, and K. K. Likharev, "Architectures for Nanoelectronic Implementation of Artificial Neural Networks: New Results", Neurocomputing 64, pp. 271-283 (2005).
7. Ö. Türel, J. H. Lee, X. Ma, and K. K. Likharev, "Nanoelectronic Neuromorphic Networks (CrossNets): New Results", in: Proc. IJCNN’04, pp. 389-394 (2004).
8. Ö. Türel, J. H. Lee, X. Ma, and K. K. Likharev, "Neuromorphic Architectures for Nanoelectronic Circuits", Int. J. of Circuit Theory and Applications 32, pp. 277-302 (2004).
9. J. H. Lee and K. K. Likharev, "CMOL CrossNets as Pattern Classifiers", Lecture Notes on Computer Science 3512, pp. 446-454 (2005).
10. J. H. Lee, X. Ma, and K. K. Likharev, "CMOL CrossNets: Possible Neuromorphic Nanoelectronic Circuits", in: Advances in Neural Information Processing Systems 18, ed. by Y. Weiss et al., MIT Press, Cambridge, MA, pp. 755-762 (2006).
11. J. H. Lee and K. K. Likharev, "In Situ Training of CMOL CrossNets", in: Proc. WCCI/IJCNN’06, pp. 5026-5034 (2006).
12. X. Ma and K. K. Likharev, "Global Reinforcement Learning in Neural Networks with Stochastic Synapses", in: Proc. WCCI/IJCNN’06, pp. 47-53 (2006).
13. X. Ma and K. K. Likharev, "Global Reinforcement Learning in Stochastic Neural Networks", IEEE Trans. on Neural Networks 18, pp. 573-577 (2007).
14. J. H. Lee and K. K. Likharev, "Defect-Tolerant Nanoelectronic Pattern Classifiers", Int. J. of Circuit Theory and Applications 35, pp. 239-264 (2007).
15. K. K. Likharev, "CrossNets: Neuromorphic Hybrid CMOS/Nanoelectronic Networks", Science of Advanced Materials 3, pp. 322-331 (2011).
16. T. J. Walls and K. K. Likharev, "Self-organization in Autonomous, Recurrent, Firing-Rate CrossNets with Quasi-Hebbian Plasticity", IEEE Trans. on Neural Networks and Learning Systems, 25, pp. 819-824 (2014).